Quote from: john117 on February 11, 2014, 11:34:16 PM

Quote from: kaene on February 11, 2014, 11:19:16 PM

Quote from: Supercomputing on February 11, 2014, 11:09:26 PM

Quote from: john117 on February 11, 2014, 11:03:16 PM

Quote from: paulthetafy on February 11, 2014, 10:55:38 PM

Quote from: john117 on February 11, 2014, 10:49:40 PM

in scanriecoin.c:209 repleace

Code:

#define MR_TESTS 12

with

Code:

#define MR_TESTS 1

and run with --max-prime=10000

speedup is about 200%
but it maybe finds more composite numbers

do you know what effect this has or are you just playing with the source?
Gatra, please can you explain what this does?

MR_TESTS is the number of miller rabin test performed to check if a number is prime,
as more tests you use the more you can be sure the number is prime.
This is important for the wallet, but not for the miner!
(if a composite is submited the network rejects is)

futur optimation coul be an implementation of the fermat test form primecoin,
which also could speed up mining a bit.

Also the miner is not optimized for cache performance!

Yes, this miner is very inefficient, I will look into it this weekend if time permits

Will this really speed up the mining process or just the speed we see in the miner? If the miner submits a composite does the wallet test it or is the network?

i just know for sure that the miner shows an higher hashrate,
but as the primecoin paper says: with bigger numbers the propability for a composite number passing the test decreases.

I don't kown if this also applays for the miller rabin test, but i just forkt gatras cpuminer
and implemented the fremat test form primecoin:

https://github.com/iddumied/cpuminer-rminerd-fermat

Nice! what's the performance gain on this?